Re: 2.15.8 Regtests

[Graham Percival]

On Sat, Aug 06, 2011 at 11:48:30PM +0000, Keith OHara wrote:
> David Kastrup <dak <at> gnu.org> writes:
> 
> > Looks like we associate differing meanings with one or several of the
> > words "unless", "somebody" or "objects".
> 
> If the test remains in input/regression, then anyone inspecting the tests
> will think that \revert of nested properties worked completely in version
> 2.14 but now fails.

Agreed; we should try to avoid giving that impression.

> Does anyone object to removing the test 'nested-property-revert.ly'
> from the input/regression?

I believe that we should (temporarily) remove
  input/regression/nested-property-revert.ly
from git master.
Note that the file itself will still be in the git history, and we
can "unearth" it using webgit (or your local git) whenever we
want.  So no need to panic about making a copy right now before it
is removed.

I believe that once we have cleared up the relevant fundamental
architecture brokenness, we should reinstate that test (or at
least, something like it).  But I think that we should treat that
"reinstatement" like a new patch, i.e. include it with whatever
programming fix patch.


I think this is exactly what Keith was proposing, but I hope that
this explanation is more clear for the email archives.

Cheers,
- Graham